Separation of physiological very low frequency fluctuation from aliasing by switched sampling interval fMRI scans

doi:10.1016/j.mri.2004.09.005

Magnetic Resonance Imaging

Volume 23, Issue 1, January 2005, Pages 41-46

https://doi.org/10.1016/j.mri.2004.09.005 Get rights and content

Abstract

Anesthetized children have dominant blood-oxygen-level-dependent (BOLD) signal sources presenting high-power fluctuations at very low frequencies (VLF <0.05 Hz). Aliasing of frequencies higher than critically sampled has been regarded as one probable origin of the VLF fluctuations. Aliased signal frequencies change when the sampling rate of the data is altered. In this study, the aliasing of VLF BOLD signal fluctuation was analysed by switching the repetition time (TR) of magnetic resonance (MR) images. Eleven anesthetized children were imaged at 1.5 T using TRs of 500 and 1200 ms. The BOLD signal sources were separated with independent component analysis (ICA). Occipital cortex signal sources had nonaliased VLF fluctuation (∼0.03 Hz) in 9 of 11 subjects. Arterial signal sources failed to present stable power peaks at frequencies lower than 0.42 Hz presumably due to aliasing. Cerebrospinal fluid (CSF)-related signal sources showed nonaliased VLF in four subjects. In conclusion, the VLF BOLD signal fluctuation in the occipital cortex is a true physiological fluctuation, not a result of signal aliasing.

Introduction

Blood-oxygen-level-dependent (BOLD) T2*-weighted magnetic resonance (MR) images reflect cortical blood flow and oxygenation alterations. The BOLD signal variability is related to either neuronal activation or spontaneous vasomotor waves [1], [2], [3], [4], [5]. The spontaneous BOLD fluctuations have a nonrandom structure and they reflect functional connectivity and default mode brain activity via vasomotor-wave-like temporal BOLD signal fluctuations [3], [5], [6], [7], [8], [9]. The very low frequency fluctuations are so strong in anesthetized child brain cortex that they can be detected as statistically independent BOLD signal sources with independent component analysis [7].

The cerebral blood flow and metabolism of children are markedly enhanced (i.e., up to two times) compared to adults [10], [11]. The elevated baseline blood flow in children may enhance the conduction of the cardiorespiratory pulses inside the cranial vault. The physiological pulsations of CSF induce steady-state free precession (SSFP) disturbances in serial echo-planar imaging (EPI) scans having shorter TR than the T2-relaxation time of CSF (T2_CSF). The SSFP disturbances can induce low-frequency signal oscillations in serial EPI image signal [12]. Physiological pulsations that have higher frequencies than critically sampled may, on the other hand, cause aliasing of fluctuation power to lower frequencies [13]. The aliasing problem is present with all conventional fMRI scans that have a long TR [13].

In children, heart pulsation may be over 2 Hz and so the image TR should be smaller than 250 ms if aliasing is to be theoretically avoided (Appendix A). Fast MRI data sampling rates and data filtering can be used to avoid signal aliasing, but then usually TR has to be set low and spatial coverage has to be compromised [13], [14], [15]. When using low TRs, the fluctuations related to disturbance of SSFP are increased and BOLD signal-to-noise ratio (SNR) is reduced [12], [13]. K-space phase data analysis and temporal reordering of images can provide data on physiological fluctuations for estimation of noise aliasing [13], [16]. The temporal reordering procedures have problems with the reordering of frequency harmonics, and spatial analysis regarding the aliased signals is not straightforward [13].

The aim of the study was to show that the VLF BOLD signal fluctuation in the occipital brain cortex of anesthetized children is not aliased noise, but rather a true physiological signal. A simple method that uses relatively high TR values for the detection of aliased signals was used. First of all, the subjects were repetitively imaged with two sampling intervals that were not a direct multiplication of the other. Secondly, the TRs chosen were rather long (i.e., 500 and 1200 ms) in order to decrease the possible physiological pulsation-induced disturbances of SSFP in serial EPI acquisitions [12]. After switching of the sampling rates, a real physiological signal fluctuation will stay in the same frequency, and aliased oscillations will shift along the frequency scale.

Section snippets

Materials and methods

The Ethical Committee of the University of Oulu had approved the study, and informed consent to the imaging was obtained from the parents. The 11 children (mean 4.9±1.4 years, eight male, three female) were imaged in order of admittance for clinical routine brain MR scans. The anesthesiologist sedated the subjects after midazolam premedication (0.3 mg/kg) with intravenous thiopental (5–6 mg/kg) boluses until the subjects slept with spontaneous breathing [5].

The imaging was performed using a

Results

No phantom scans showed any sign of elevated low-frequency fluctuations (Fig. 1). The 40 ICA signal sources of the phantom data showed no signs of signal sources typical of the ones detected in the subjects in spatial or frequency domain. None of the studied subjects presented significant motion (>0.6 mm) artefacts in the center of mass (COM) analysis. In the total image spectrum, the physiological signal-related power peaks are elevated clearly over the noise level typical of phantom data (

Discussion

The prominent BOLD signal sources in the occipital cortices of the anesthetized children associated with spontaneous VLF fluctuations are not aliased artefacts. Nine of 11 subjects showed stable VLF power peaks in the occipital cortex of the two consecutive scans. The VLF fluctuation in these sources originated from real physiological phenomenon rather than from aliased noise. The SSFP fluctuations are present when the TR used is less than the T2 relaxation time of CSF. However, the VLF BOLD

Conclusion

The statistically independent BOLD signal sources presenting very low frequency (VLF) fluctuation in the brain occipital cortex of the anesthetized children are physiological fluctuations and not aliased noise. Arterial BOLD signal sources are affected by aliasing. Switched sampling interval scanning enables the separation of aliased and nonaliased BOLD sources with correctly selected fMRI image sampling rates.

References (27)

V. Kiviniemi et al.
Independent component analysis of non-deterministic fMRI signal sources
Neuroimage
(2003)
S. Ogawa et al.
Brain magnetic resonance imaging with contrast dependent on blood oxygenation
Proc. Natl. Acad. Sci.
(1990)
S. Ogawa et al.
On the characteristics of functional magnetic resonance imaging of the brain
Annu. Rev. Biophys. Biomol. Struct.
(1998)
B.B. Biswal et al.
Functional connectivity in the motor cortex of resting human brain using echo-planar MRI
Magn. Reson. Med.
(1995)
P.P. Mitra et al.
The nature of spatiotemporal changes in cerebral hemodynamics as manifested in functional magnetic resonance imaging
Magn. Reson. Med.
(1997)
V. Kiviniemi et al.
Slow vasomotor fluctuation in the fMRI of the anesthetized child brain
Magn. Reson. Med.
(2000)
S.-J. Li et al.
Alzheimer disease: evaluation of functional MR imaging index as a marker
Radiology
(2002)
M.D. Greicius et al.
Functional connectivity on the resting brain: a network analysis of the default mode hypothesis
Proc. Natl. Acad. Sci.
(2003)
H. Laufs et al.
Electroencephalographic signatures of attentional and cognitive default modes in spontaneous brain activity fluctuations at rest
Proc. Natl. Acad. Sci.
(2003)
H.T. Chugani et al.
Maturational changes in cerebral function in infants determined by 18FDG positron emission tomography
Science
(1986)

T. Takahashi et al.

Developmental changes of cerebral blood flow and oxygen metabolism in children

AJNR. Am. J. Neuroradiol.

(1999)

X. Zhao et al.

B₀-fluctuation-induced temporal variation in EPI image series due to disturbance of steady-state free precession

Magn. Reson. Med.

(2000)

L. Frank et al.

Estimation of respiration-induced noise fluctuations from undersampled multislice fMRI data

Magn. Reson. Med.

(2001)

Cited by (35)

Ultra-slow Oscillations in fMRI and Resting-State Connectivity: Neuronal and Vascular Contributions and Technical Confounds
2020, Neuron
Citation Excerpt :
These difficulties stand in contrast to the significant efforts made for human BOLD fMRI studies (Birn et al., 2006; Dagli et al., 1999; Glover et al., 2000; Hu et al., 1995; Shmueli et al., 2007; Wise et al., 2004). The adoption of signal processing methods developed for human studies (Caballero-Gaudes and Reynolds, 2017; Murphy et al., 2013) will be crucial to reduce the interference of cardiorespiratory artifacts with the spatiotemporal patterns of the resting-state fMRI signals in animals (Abreu et al., 2017; Biswal et al., 1996; Bright and Murphy, 2017; Feinberg and Setsompop, 2013; Kiviniemi et al., 2005). MRI acquisition relies on Larmor frequency-based spatial encoding through the sequence-specific magnetic field gradients across a subject.
Ultra-slow, ∼0.1-Hz variations in the oxygenation level of brain blood are widely used as an fMRI-based surrogate of “resting-state” neuronal activity. The temporal correlations among these fluctuations across the brain are interpreted as “functional connections” for maps and neurological diagnostics. Ultra-slow variations in oxygenation follow a cascade. First, they closely track changes in arteriole diameter. Second, interpretable functional connections arise when the ultra-slow changes in amplitude of γ-band neuronal oscillations, which are shared across even far-flung but synaptically connected brain regions, entrain the ∼0.1-Hz vasomotor oscillation in diameter of local arterioles. Significant confounds to estimates of functional connectivity arise from residual vasomotor activity as well as arteriole dynamics driven by self-generated movements and subcortical common modulatory inputs. Last, methodological limitations of fMRI can lead to spurious functional connections. The neuronal generator of ultra-slow variations in γ-band amplitude, including that associated with self-generated movements, remains an open issue.
Combined spatiotemporal ICA (stICA) for continuous and dynamic lag structure analysis of MREG data
2017, NeuroImage
Citation Excerpt :
This is not a trivial matter since the brain was shown to present strong cardiorespiratory pulsations and ubiquitous very low frequency (VLF < 0.1 Hz) vasomotor waves propagate spatially over the whole brain. The cardiorespiratory pulsations aliases in classical TR > 400 ms BOLD signal (Kiviniemi et al., 2005; Kiviniemi et al., 2015; Majeed et al., 2011; Thompson et al., 2014; Thompson et al., 2015). Non-critically sampled pulsations in different frequencies also mix with natural motion artifacts, which will distort time signals in a unique pattern in each voxel and produce virtually un-recoverable signal changes in 2D echo planar imaging (EPI) data (Beall and Lowe, 2014).
This study investigated lag structure in the resting-state fMRI by applying a novel independent component (ICA) method to magnetic resonance encephalography (MREG) data. Briefly, the spatial ICA (sICA) was used for defining the frontal and back nodes of the default mode network (DMN), and the temporal ICA (tICA), which is enabled by the high temporal resolution of MREG (TR=100ms), was used to separate both neuronal and physiological components of these two spatial map regions. Subsequently, lag structure was investigated between the frontal (DMNvmpf) and posterior (DMNpcc) DMN nodes using both conventional method with all-time points and a sliding-window approach.
A rigorous noise exclusion criterion was applied for tICs to remove physiological pulsations, motion and system artefacts. All the de-noised tICs were used to calculate the null-distributions both for expected lag variability over time and over subjects. Lag analysis was done for the three highest correlating denoised tICA pairs.
Mean time lag of 0.6 s (± 0.5 std) and mean absolute correlation of 0.69 (± 0.08) between the highest correlating tICA pairs of DMN nodes was observed throughout the whole analyzed period. In dynamic 2 min window analysis, there was large variability over subjects as ranging between 1–10 sec. Directionality varied between these highly correlating sources an average 28.8% of the possible number of direction changes.
The null models show highly consistent correlation and lag structure between DMN nodes both in continuous and dynamic analysis. The mean time lag of a null-model over time between all denoised DMN nodes was 0.0 s and, thus the probability of having either DMNpcc or DMNvmpf as a preceding component is near equal. All the lag values of highest correlating tICA pairs over subjects lie within the standard deviation range of a null-model in whole time window analysis, supporting the earlier findings that there is a consistent temporal lag structure across groups of individuals. However, in dynamic analysis, there are lag values exceeding the threshold of significance of a null-model meaning that there might be biologically meaningful variation in this measure. Taken together the variability in lag and the presence of high activity peaks during strong connectivity indicate that individual avalanches may play an important role in defining dynamic independence in resting state connectivity within networks.
Resting-state fMRI confounds and cleanup
2013, NeuroImage
Citation Excerpt :
Accounting for non-neural noise by using high sampling rates was popular in the early days of resting-state fMRI (Biswal et al., 1995; Cordes et al., 2001; De Luca et al., 2006; Lowe et al., 1998). Kiviniemi and colleagues, taking advantage of this aliasing of physiological noise, demonstrated that low-frequency BOLD signal fluctuations exist that are unrelated to the aliased cardiac and respiratory frequencies (Kiviniemi et al., 2005). One drawback of low-TR sampling is limited spatial coverage.
The goal of resting-state functional magnetic resonance imaging (fMRI) is to investigate the brain's functional connections by using the temporal similarity between blood oxygenation level dependent (BOLD) signals in different regions of the brain “at rest” as an indicator of synchronous neural activity. Since this measure relies on the temporal correlation of fMRI signal changes between different parts of the brain, any non-neural activity-related process that affects the signals will influence the measure of functional connectivity, yielding spurious results. To understand the sources of these resting-state fMRI confounds, this article describes the origins of the BOLD signal in terms of MR physics and cerebral physiology. Potential confounds arising from motion, cardiac and respiratory cycles, arterial CO₂ concentration, blood pressure/cerebral autoregulation, and vasomotion are discussed. Two classes of techniques to remove confounds from resting-state BOLD time series are reviewed: 1) those utilising external recordings of physiology and 2) data-based cleanup methods that only use the resting-state fMRI data itself. Further methods that remove noise from functional connectivity measures at a group level are also discussed. For successful interpretation of resting-state fMRI comparisons and results, noise cleanup is an often over-looked but essential step in the analysis pipeline.
The role of physiological noise in resting-state functional connectivity
2012, NeuroImage
Citation Excerpt :
First, correlations are observed between functionally related brain areas. Second, early work on resting-state connectivity showed that correlated signal fluctuations in functionally related brain regions are not due to cardiac pulsations and respiration but instead occur at much lower temporal frequencies (Biswal et al., 1995; Cordes et al., 2001; Kiviniemi et al., 2005). Third, these fluctuations have similar characteristics as neuronally-induced BOLD responses (e.g. TE dependence, attenuation of the response during hypercapnia) (Biswal et al., 1997; Peltier and Noll, 2002).
Functional connectivity between different brain regions can be estimated from MRI data by computing the temporal correlation of low frequency (< 0.1 Hz) fluctuations in the MRI signal. These correlated fluctuations occur even when the subject is “at rest” (not asked to perform any particular task) and result from spontaneous neuronal activity synchronized within multiple distinct networks of brain regions. This estimate of connectivity, however, can be influenced by physiological noise, such as cardiac and respiratory fluctuations. This brief review looks at the effect of physiological noise on estimates of resting-state functional connectivity, discusses ways to remove physiological noise, and provides a personal recollection of the early developments in these approaches. This review also discusses the importance of physiological noise correction and provides a summary of evidence demonstrating that functional connectivity does have a neuronal underpinning and cannot purely be the result of physiological noise.
Functional connectivity MRI in infants: Exploration of the functional organization of the developing brain
2011, NeuroImage
Citation Excerpt :
Beginning with Biswal's landmark observation of correlated functional activation at frequencies less than 0.08 Hz (Biswal et al., 1995), exploration of RSNs has centered on low frequency bands. While initial investigations demonstrated low-frequency BOLD fluctuations resulted from neuronal activity rather than physiological artifact (Lowe et al., 1998; Cordes et al., 2001; Kiviniemi et al., 2005a; De Luca et al., 2006; Wu et al., 2008), subsequent studies illustrated functional connections exhibit increasing power at lower frequencies and further defined spectral features of specific RSNs (Cordes et al., 2001; Achard et al., 2006; Fox and Raichle, 2007; Salvador et al., 2008; Wu et al., 2008). Connection distance was found to play an important role in frequency specificity, with long distance connections appearing more frequency specific (peaking at lower frequencies) than shorter connections.
Advanced neuroimaging techniques have been increasingly applied to the study of preterm and term infants in an effort to further define the functional cerebral architecture of the developing brain. Despite improved understanding of the complex relationship between structure and function obtained through these investigations, significant questions remain regarding the nature, location, and timing of the maturational changes which occur during early development. Functional connectivity magnetic resonance imaging (fcMRI) utilizes spontaneous, low frequency (< 0.1 Hz), coherent fluctuations in blood oxygen level dependent (BOLD) signal to identify networks of functional cerebral connections. Due to the intrinsic characteristics of its image acquisition and analysis, fcMRI offers a novel neuroimaging approach well suited to investigation of infants. Recently, this methodology has been successfully applied to examine neonatal populations, defining normative patterns of large-scale neural network development in the maturing brain. The resting-state networks (RSNs) identified in these studies reflect the evolving cerebral structural architecture, presumably driven by varied genetic and environmental influences. Principal features of these investigations and their role in characterization of the tenets of neural network development during this critical developmental period are highlighted in this review. Despite these successes, optimal methods for fcMRI data acquisition and analysis for this population have not yet been defined. Further, appropriate schemes for interpretation and translation of fcMRI results remain unknown, a matter of increasing importance as functional neuroimaging findings are progressively applied in the clinical arena. Notwithstanding these concerns, fcMRI provides insight into the earliest forms of cerebral connectivity and therefore holds great promise for future neurodevelopmental investigations.
Effects of heartbeat and respiration on macaque fMRI: Implications for functional connectivity
2010, Neuropsychologia
Citation Excerpt :
It is important to note that these differences are not an intrinsic property of the two signals, because the observed wavelength depends on the aliasing of the artifacts. The aliased frequency is determined by the difference between TR and the average duration and variability of the cardiac or respiratory cycle (e.g., Glover et al., 2000; Kiviniemi, Ruohonen, & Tervonen, 2005; Lund, 2001a). Aliasing into low frequency bands is especially likely to occur if the TR and duration of the artifactual cycle are similar.
The use of functional magnetic resonance imaging (fMRI) in non-human primates is on the increase. It is known that the blood-oxygen-level-dependent (BOLD) signal varies not only as a function of local neuronal energy consumption but also as a function of cardiac and respiratory activity. We mapped these cyclic cardiac and respiratory artifacts in anesthetized macaque monkeys and present an objective analysis of their impact on estimates of functional connectivity (fcMRI). Voxels with significant cardiac and respiratory artifacts were found in much the same regions as previously reported for awake humans. We show two example seeds where removing the artifacts clearly decreased the number of false positive and false negative correlations. In particular, removing the artifacts reduced correlations in the so-called resting state network. Temporal bandpass filtering or spatial smoothing may help to reduce the effects of artifacts in some cases but are not an adequate replacement for an algorithm that explicitly models and removes cyclic cardiac and respiratory artifacts.

View all citing articles on Scopus

View full text

Separation of physiological very low frequency fluctuation from aliasing by switched sampling interval fMRI scans

Abstract

Introduction

Section snippets

Materials and methods

Results

Discussion

Conclusion

Neuroimage

Brain magnetic resonance imaging with contrast dependent on blood oxygenation

Proc. Natl. Acad. Sci.

On the characteristics of functional magnetic resonance imaging of the brain

Annu. Rev. Biophys. Biomol. Struct.

Functional connectivity in the motor cortex of resting human brain using echo-planar MRI

Magn. Reson. Med.

The nature of spatiotemporal changes in cerebral hemodynamics as manifested in functional magnetic resonance imaging

Magn. Reson. Med.

Slow vasomotor fluctuation in the fMRI of the anesthetized child brain

Magn. Reson. Med.

Alzheimer disease: evaluation of functional MR imaging index as a marker

Radiology

Functional connectivity on the resting brain: a network analysis of the default mode hypothesis

Proc. Natl. Acad. Sci.

Electroencephalographic signatures of attentional and cognitive default modes in spontaneous brain activity fluctuations at rest

Proc. Natl. Acad. Sci.

Maturational changes in cerebral function in infants determined by 18FDG positron emission tomography

Science

Developmental changes of cerebral blood flow and oxygen metabolism in children

AJNR. Am. J. Neuroradiol.

B0-fluctuation-induced temporal variation in EPI image series due to disturbance of steady-state free precession

Magn. Reson. Med.

Estimation of respiration-induced noise fluctuations from undersampled multislice fMRI data

Magn. Reson. Med.

B₀-fluctuation-induced temporal variation in EPI image series due to disturbance of steady-state free precession